AI demand to push memory fab spending past USD $50bn

Worldwide 300mm fab equipment investment in the memory sector is projected to reach USD $52 billion in 2026, according to SEMI. If achieved, it would be the first time spending has exceeded USD $50 billion.

SEMI expects investment to rise a further 11% to USD $57 billion in 2027, based on its latest 300mm Fab Outlook. Over the 2024 to 2029 period, the industry body forecasts a compound annual growth rate of 19% for memory-related 300mm fab equipment spending.

The revised outlook points to heavier spending on advanced memory used in artificial intelligence systems, particularly high bandwidth memory, DDR5 and 3D NAND. These products are being deployed across AI infrastructure, data centres and newer computing systems, increasing demand for both production tools and manufacturing capacity.

Segment growth

DRAM is expected to account for the larger share of the increase. SEMI projects DRAM equipment spending will climb 29% to USD $37 billion in 2026, supported by demand for HBM and DDR5 used alongside graphics processing units and other AI accelerators.

Spending on 3D NAND equipment is forecast to rise 28% to USD $14 billion in 2026. The report attributes that increase to higher storage requirements as AI deployment expands.

The capacity outlook has also been revised higher. Worldwide 300mm memory capacity is projected to reach 4.1 million wafers per month in 2026 and 4.2 million wafers per month in 2027.

The increase comes as manufacturers continue to invest in advanced-node DRAM and higher-layer 3D NAND. At the same time, effective capacity growth remains constrained by technology migration and more complex production processes, including transitions tied to advanced-node DRAM, HBM and higher-layer NAND.

Ajit Manocha, President and Chief Executive Officer of SEMI, said the spending pattern reflects a broader shift in chip industry priorities as artificial intelligence workloads drive demand for newer memory technologies.

"Strong demand for high bandwidth memory and other advanced memory technologies is reshaping investment priorities across the semiconductor supply chain," said Ajit Manocha, President and Chief Executive Officer of SEMI.

He said the build-out of AI systems was influencing both capacity additions and process upgrades at memory manufacturers.

"As AI infrastructure expands, memory manufacturers are accelerating investments in both capacity and technology migration to support the next wave of data-intensive applications," said Manocha.

Wider market

The latest edition of the 300mm Fab Outlook includes a higher projection for memory-sector equipment spending than the previous version. SEMI said the increase was driven by rising capital expenditure plans among leading cloud service providers and strong demand for AI accelerators.

The report tracks 413 facilities and lines globally. It includes 155 updates and seven new fab or line projects since the previous publication.

The figures underline how memory has become a central part of the semiconductor investment cycle as AI-related infrastructure spending broadens beyond processors to the components needed to feed them with data. High bandwidth memory in particular has become a focal point for manufacturers and equipment suppliers as chipmakers race to meet demand from AI systems.

While logic chips have drawn much of the attention around AI, the latest forecast suggests memory producers are entering a similarly heavy investment phase. Spending on memory-related 300mm fab equipment is projected to approach USD $80 billion by 2029, placing the segment among the semiconductor industry's largest areas of capital expenditure.